The fastest voice to text for Mac. Private by design.
Hold a key, speak, and your words appear in any app, instantly. By default, speech recognition runs on your Mac.
macOS 14+ · Apple Silicon · Free download
Three steps, and no new habits to learn.
Hold the key
Press and hold Right Option in any app. You can pick a different shortcut in Settings.
Speak naturally
Talk the way you normally would. The pill at the bottom of your screen shows WhisperWrite is listening.
Let go, it's typed
Release the key. Your words appear where your cursor is, punctuated and without the ums.
Prefer hands-free? Tap the key once to start and tap again to finish. Esc cancels.
Text in 0.1 seconds.
With the on-device engine, the default, WhisperWrite runs speech recognition right on your Mac, so there's no upload and no waiting on a server. On an M1 Pro, it transcribes a sentence in about a tenth of a second.
It's built on NVIDIA's Parakeet family of open speech models, which don't invent words from silence or background noise.
By default, your voice never leaves your Mac.
With the on-device engine, the default, speech recognition happens on your computer, not on a server, and your recordings are never uploaded. There's nothing to sign up for.
On-device
By default, speech recognition runs entirely on your Mac.
Works offline
After a one-time model download, dictate on a plane, on a train or with Wi-Fi turned off.
No account
Download, open and start talking. No sign-up, no email address.
Audio stays local
With the on-device engine, your recordings are never sent off your computer.
The OpenAI speech engine is optional and off by default. If you choose it on the Speech Model page and add your own OpenAI API key, each recording is sent over HTTPS to OpenAI to be transcribed, along with your language and dictionary words. It goes straight from your Mac to OpenAI. We never receive it. See exactly what's sent
AI Edit is optional and off by default. It uses your own Anthropic API key. When it's on, your transcribed text is sent to Claude for polishing, along with your dictionary words, your writing style and the name of the app you're in. AI Edit never sends your audio. See exactly what's sent
Everything you need to write by talking.
Works in every app
Mail, Slack, Notes, Google Docs, ChatGPT, VS Code. If there's a cursor, WhisperWrite can type there.
Cleans up ums
Removes "um" and "uh", adds punctuation and understands "new line" and "new paragraph".
Dictionary and snippets
Teach it names and jargon. Say "my email" and it types your email address.
AI Edit
Optional polish with Claude, using your own Anthropic API key. Fixes grammar, applies "actually, I meant…" corrections, formats lists and matches the tone of the app you're in.
Command Mode
Select text anywhere, hold a shortcut and say "make this friendlier" or "turn this into bullet points". Uses your AI Edit key.
Hands-free mode
Tap the key to start and tap again to finish, or click Record in the menu bar. Press Esc to cancel at any time.
History and stats
See words dictated, speaking speed, time saved and your day streak. Search everything you've said.
Paste last dictation
Press Control-Command-V to paste your last dictation again, in any app.
See how much you've said, and teach it your words.
Yes. With the on-device engine, the default, speech recognition runs entirely on your Mac, and your audio is never saved or sent anywhere. There's no account to create. WhisperWrite checks for updates and shares basic install statistics, which you can turn off in Settings. If you choose the optional OpenAI speech engine, your recordings go to OpenAI using your own API key. If you turn on AI Edit, text (never audio) goes to Anthropic using your own API key. The Privacy Policy has the details.
Start talking. Your Mac will keep up.
Free download · macOS 14+ · Apple Silicon · Signed and notarized by Apple